Mineria De Datos

white paper

Data Mining:
an introduction
Clementine™ – Working with health care

white paper

Data Mining: an introduction

2

Abstract
Health care and pharmaceutical companies face an explosion of data arising from
clinical, administrative, commercial and scientific activities. There are many traditional
techniques for analyzing data including statistics, management reporting anddata
display. Data mining offers a new approach to data analysis using techniques based on
machine learning (algorithms derived from research in artificial intelligence), alongside
the conventional methods.
These techniques work by “learning” patterns in data. They find patterns and make
predictions, which elude all but the most expert users of conventional methods.
In addition, they generatedecision or prediction models, based on the actual historical
data. These models are synthesized — not programmed explicitly by a programmer or
physician. Thus, they represent true evidence-based decision support.
Health care and pharmaceutical professionals have a special duty of care, as their
decisions may be a matter of life and death for their clients. The esoteric nature of data
miningcan distance these professionals from the models. This paper discusses how
data mining can be packaged in such a way that professionals can take part directly
in data mining and thus assess the implications of using data mining in safety-critical or
service-critical applications.
The data flood
The Proceedings of the 1995 Conference in Knowledge Discovery in Databases (KDD)
opens with thefollowing quotations:
“It is estimated that the amount of information in the world doubles every 20 months.
What are we supposed to do with this flood of raw data? Clearly little of it will ever be
seen by human eyes.”
“Computers promised fountains of wisdom but delivered floods of data.”
You are familiar with this phenomenon of data explosion. Computerized systems collect
data about a myriad ofeveryday transactions: at supermarket checkouts, bank cash
machines, airline tickets, phone calls, buying gasoline, the list is endless.
Health care providers, insurers and suppliers, and the pharmaceutical industries
contribute their share
s Administrative systems log patient admissions and discharges, resource utilization in
hospitals and practices, delivery and use of supplies, staffshift patterns and hours
worked, costs of procedures and times taken for procedures.
s Every clinical act and its outcome are recorded. Patient records are being transferred
to electronic form.
s Pharmaceutical knowledge increases daily, with new compounds, new dosage patterns,
etc. Clinical trials generate huge amounts of data which must be analyzed and which,
because of the controlledenvironment in which they are collected, should be an
invaluable resource for future study.

white paper

Data Mining: an introduction

s

s

3

Marketing and sales by pharmaceutical manufacturers and pharmacies provide the
return on the huge investments required to bring new drugs to the market. Every such
sale can be captured as data and analyzed to help make more sales and predictfuture
demand.
Scientists in companies and hospitals gather mountains of experimental and
laboratory data.

Information Systems managers are only too aware of this data explosion. They continually
have to upgrade computers with more disk storage. The IT industry bombards them with
offers of the latest databases, data warehouses, data marts and a host of data translation,
transformation andreporting tools all aiming to tame the data explosion.
Data holds knowledge
Data holds the record of the organization’s performance in all of its business areas.
A hospital, which has been admitting patients for many years, has data it can use to
estimate accurately the likely cost of treatment and length of stay of a patient. A health
care insurer has data about all of its subscribers,...